# Initialization and log probability problem

**URL:** <https://discourse.mc-stan.org/t/initialization-and-log-probability-problem/14091>\
**Category:** RStan\
**Created:** [April 5, 2020, 9:20am UTC](https://discourse.mc-stan.org/t/initialization-and-log-probability-problem/14091 "2020-04-05T09:20:38Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![FU74](https://avatars.discourse-cdn.com/v4/letter/f/57b2e6/32.png) [@FU74](https://discourse.mc-stan.org/u/FU74)\
**Post date:** [April 5, 2020, 9:20am UTC](https://discourse.mc-stan.org/t/initialization-and-log-probability-problem/14091/1 "2020-04-05T09:20:39Z")

</div>

I have following code which give me error message:

Chain 1: Log probability evaluates to log(0), i.e. negative infinity.  
Chain 1: Stan can’t start sampling from this initial value.  
Chain 1:  
Chain 1: Initialization between (-2, 2) failed after 100 attempts.  
Chain 1: Try specifying initial values, reducing ranges of constrained values, or reparameterizing the model.  
[1] “Error in sampler$call\_sampler(args\_list[[i]]) : Initialization failed.”  
[1] “error occurred during calling the sampler; sampling not done”

The R-code is:

dat=read.table(“Loss Reserve Data.txt”); st=1;  
N=dim(dat)[1] #number of rows  
colnames(dat)=c(“Year”,“Y”,“poli”,“develop”)  
y0=0  
y=dat[,2]

burn=1000  
burn1=burn+1  
iters=10000

dat$poli\<-factor(dat$poli)  
dat$develop\<-factor(dat$develop)

poli=as.integer(dat$poli) # rows i effect (policy year)  
develop=as.integer(dat$develop) # columns j effect (development year)

I=nlevels(dat$poli);I1=I-1; #number of levels of policy (i)  
J=nlevels(dat$develop);J1=J-1 #number of levels of development (j)

P=I1+6 #no of par in the model

#begin Rstan program

trialmodelA\<- "

//Generalized Beta 2 Distribution

functions{  
real beta2\_log(real x, real a, real mu, real p, real q) {  
return log(fabs(a))+(a\*p-1)_log(x)+lgamma(p+q)+(a_p)_lgamma(p+(1/a))+(a_p)_lgamma(q-(1/a))-(a_p)_log(mu)-(a_p)_lgamma§-(a_p)\*lgamma(q)-lgamma§-lgamma(q)-(p+q)_log(1+((x_tgamma(p+(1/a))_tgamma(q-(1/a)))/(mu_tgamma§\*tgamma(q)))^a);  
}  
}  
//data block  
data {  
int\<lower=1\> N; // number of observations  
real y[N]; // amounts of claims  
real y0;  
int\<lower=1\> I; //number of policy (i)  
int\<lower=1\> J; //number of develop (j)  
int\<lower=1\> I1; //dimension of alpha  
int\<lower=1\> J1; //dimension of beta  
int\<lower=1, upper=I\> poli[N]; //policy id  
int\<lower=1, upper=J\> develop[N]; //develop id  
}

parameters {  
real mu0; //mean effect  
real alpha[I1];  
real gamma1;  
real gamma2;  
real a;  
real p;  
real q;  
}

transformed parameters {  
real mu[N];  
real alpha18;

alpha18=-sum(alpha[1:I1]);

for (i in 1:N) {  
if (develop[i]==1 && poli[i]!=18) {mu[i]=exp(mu0+alpha[poli[i]]+gamma1_y0);} //case A with y0  
if (develop[i]==18) {mu[i]=exp(mu0+alpha[poli[i]]+gamma1_y[i-1]+gamma2_mu[i-1]);} //case B with beta18  
if (poli[i]==18) {mu[i]=exp(mu0+alpha18+gamma1_y0);} //case C with alpha18  
if (develop[i]!=1 && develop[i]!=18) {mu[i]=exp(mu0+alpha[poli[i]]+gamma1_y[i-1]+gamma2_mu[i-1]);} //else  
}  
}

model {  
mu0~ gamma(1,1);  
alpha[1:I1] ~ normal(0,1);  
gamma1 ~ gamma(1,1);  
gamma2 ~ gamma(1,1);  
a ~ gamma(1,1);  
p ~ gamma(1,1);  
q ~ gamma(1,1);

for (i in 1:N){  
y[i]~beta2(a, mu[i], p, q);  
}  
}

generated quantities {  
real loglik[N];  
real dev;

for (i in 1:N){  
loglik[i]=log(fabs(a))+(a\*p-1)_log(y[i])+lgamma(p+q)+(a_p)_lgamma(p+(1/a))+(a_p)_lgamma(q-(1/a))-(a_p)_log(mu[i])-(a_p)_lgamma§-(a_p)\*lgamma(q)-lgamma§-lgamma(q)-(p+q)\*log(1+((y[i]\*tgamma(p+(1/a))\*tgamma(q-(1/a)))/(mu[i]\*tgamma§\*tgamma(q)))^a);  
}

dev = -2\*sum(loglik[]);  
}  
"  
datt = list(N=N,y=y,y0=y0,I=I,J=J,I1=I1,J1=J1,poli=poli,develop=develop)

fitA = stan(model\_code = trialmodelA,  
model\_name = “Trial modelA”,  
data = datt, iter = iters,  
warmup=burn,  
chains = 1)

The data is attached.[Loss Reserve Data.txt](https://discourse.mc-stan.org/uploads/short-url/hzwRmQfdt8KRkA2vxFKGSDNZW6c.txt) (2.6 KB)

---

<div class="post-metadata">

**Author:** ![andre.pfeuffer](https://avatars.discourse-cdn.com/v4/letter/a/e480ec/32.png) [@andre.pfeuffer](https://discourse.mc-stan.org/u/andre.pfeuffer)\
**Post date:** [April 8, 2020, 7:01pm UTC](https://discourse.mc-stan.org/t/initialization-and-log-probability-problem/14091/2 "2020-04-08T19:01:17Z")

</div>

```
real gamma1;
real gamma2;
real a;
real p;
real q;

```

show all the constrainted by `lower=0`. Are you sure non in of the mu[i] cases in `for(i in 1:N) ...` loop is not assigned?

Besides that, please reformat your posting. It is really hard to follow.